|
Activity Number:
|
7
- Advances in Multivariate Spatial Process Modeling for Environmental Data
|
|
Type:
|
Invited
|
|
Date/Time:
|
Sunday, August 7, 2022 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistics and the Environment
|
|
Abstract #320419
|
|
|
Title:
|
Multivariate Localization Functions for Use in Strongly Coupled Data Assimilation
|
|
Author(s):
|
Zofia Claire Stanley* and Ian Grooms and Will Kleiber
|
|
Companies:
|
CIRES/NOAA PSL and University of Colorado, Boulder and University of Colorado
|
|
Keywords:
|
data assimilation;
spatial statistics;
multivariate;
localization;
covariance tapering
|
|
Abstract:
|
Weather forecasts fuse real world observations and numerical models in a process called data assimilation. Covariance tapering is used in data assimilation, under the name "localization", to mitigate the impact of sampling-errors in empirical covariance matrices. In atmosphere-ocean coupled models the correlation length scale varies widely and covariance tapering must account for distinct length scales of different processes. We present a new compactly supported multivariate correlation function which extends the fifth-order piecewise rational Gaspari-Cohn correlation function. We compare different localization functions in a simple data assimilation scheme on a coupled Lorenz 96 model, and illustrate that the new multivariate covariance structure improves assimilation over an array of existing models. The new multivariate model can additionally be used for likelihood calculations or kriging of large systems of multivariate data via tapering.
|
Authors who are presenting talks have a * after their name.
